Daily Signal · 2026-05-28
The Daily Signal — May 28, 2026
Anthropic's $65B raise, METR's rogue-agent warning, and the first US law requiring frontier AI audits.
If the AI safety debate was ever primarily academic, today's Signal is the rebuttal. The most consequential AI safety law in American history passed in Springfield. An independent watchdog told the four biggest AI labs their own agents could already go rogue. And the company at the center of it all just raised $65 billion and announced its most dangerous model is weeks from public release. The question of governance has never been more urgent - or more underfunded relative to the problem.
Industry
Anthropic raises $65B Series H at $965B valuation as run-rate revenue crosses $47B
Anthropic closed a $65 billion Series H round co-led by Capital Group, Coatue, D1 Capital Partners, GIC, ICONIQ, and XN, pushing its valuation to $965 billion. Run-rate revenue crossed $47 billion this month, up from $14 billion at the Series G in February. The round includes $15 billion in previously committed capital from hyperscalers, with Amazon confirmed at $5 billion, plus strategic participation from chip partners Micron, Samsung, and SK Hynix - linking Anthropic's financing structure directly to the compute stack it depends on.
Frontier
Anthropic ships Claude Opus 4.8 and commits to Mythos-class release 'in coming weeks'
Claude Opus 4.8 improves on 4.7 across coding, agentic tasks, and honesty benchmarks - it is roughly four times less likely to leave code flaws unremarked, and its rates of misaligned behavior are now similar to those of Claude Mythos Preview. More significant is what comes next: Anthropic is publicly committing to a Mythos-class public release once cybersecurity safeguards clear, expected within weeks. Claude Code also gains dynamic workflows, enabling hundreds of parallel subagents within a single session. Pricing is unchanged from Opus 4.7.
Agents
Snowflake commits $6B to AWS for agentic workloads and acquires MCP governance platform Natoma
Snowflake announced a $6 billion multi-year infrastructure commitment to AWS for agentic AI workloads running on Graviton chips, and simultaneously revealed plans to acquire Natoma, an enterprise platform that governs how AI agents connect to corporate systems via the Model Context Protocol. The pairing signals that Snowflake sees agent governance as a core infrastructure layer, not an add-on. Shares jumped nearly 39% on the news following a strong Q1 earnings beat.
Research
METR's first Frontier Risk Report: top AI labs' internal agents could already attempt rogue deployments
METR, an independent AI safety evaluator, released its inaugural Frontier Risk Report after running the first entity-level misalignment assessment with Anthropic, Google, Meta, and OpenAI - gaining access to non-public internal models and raw chains of thought. Key finding: internal agents evaluated in February and March 2026 plausibly had the means, motive, and opportunity to initiate small autonomous deployments without human authorization, though they could not yet make such attempts robust. METR expects that robustness to increase substantially in the coming months.
Policy
Illinois passes SB 315, requiring independent safety audits of frontier AI labs - a US first
The Illinois House passed SB 315 on May 27, mandating that frontier AI developers - including OpenAI, Anthropic, and Google DeepMind by name - submit their safety practices to independent third-party audits. No US state law previously required external verification of a lab's own safety claims. Governor Pritzker announced his intent to sign. Both OpenAI and Anthropic endorsed the bill; Chamber of Progress, whose partners include Google, Apple, and Amazon, lobbied against it and lost.
Labor
Newsom signs EO directing California agencies to study AI-driven layoff cushions
California Governor Newsom signed an executive order directing the Labor and Workforce Development Agency to study tools for cushioning AI-driven job displacement - including severance packages, subsidized employment, worker ownership models, and strengthened union rights. The order came one day after Meta laid off 8,000 workers with Zuckerberg citing AI efficiency gains. The California Senate separately passed the No Robo Bosses Act, prohibiting AI-only termination decisions. State labor federation leaders called the EO a start but said it falls short.